API Cloud Performance Analysis Report

A Report from APImetrics & API.expert

Dr Paul M Cray


APImetrics made more than 187 million API calls to more than 4200 API endpoints in 2019 and more than 208 million API calls to more than 4600 API endpoints in 2020, in both cases from 80 cloud data center locations for AWS, Azure, Google, and IBM.

This report builds on our unrivaled collection of historical API data to establish a baseline for API quality and performance. The report focuses on a standardized selection of data from leading services, including those from prominent corporate infrastructure providers, financial services institutions, social networks, and search engines. This has allowed us to provide a uniquely comprehensive analysis of latency and availability across multiple clouds, data centers, and regions from January 2019 through December 2020.

The report also captures the effect of the massive digital transformation brought about by the COVID-19 pandemic.

Our key finding is that, across all clouds, latency decreased between the beginning of 2019 and the middle of 2020, indicating a general improvement in API performance. It then increases sharply. We believe this increase in latency to be due to significantly increased load from significant growth in global digital operations as the COVID-19 pandemic has continued. We can speculate that in the early stages of the pandemic (April 2021-July 2021), businesses were not yet fully operationally digitally. Still, in the period post-summer (Northern Hemisphere 2021), there has been a significant increase in online operations and the use of APIs, especially from regions outside of North America.

We're not planning on naming and shaming here. However, we do want to recognize the top performers of 2020. Overall, for availability, the highest performer when it comes to pass rates: was DocuSign. From January 2019 through December 2020, DocuSign had zero measurable downtime on their APIs. This is a remarkable achievement, and we send hearty congratulations to the DevOps team and everyone else at DocuSign involved in such an achievement.

The next best performer had a failure equivalent of just 18.5 minutes of downtime over 24 months, which is still a pretty good effort. To put that in perspective, the worst-performing API had over five and a half days of downtime. That's using the crudest measure we have.

While, generally, API services showed only minor slowdowns regarding the dramatic change in working habits brought on by COVID-19, one service had a small impact early on but soon recovered. The provider was Slack, and they deserve special mention for the minor impact of such a dramatic change and the performance recovery they made almost immediately.

Suppose you wish to explore the datasets behind this report. In that case, you can do so at API.expert, where for a small monthly fee, you will be able to see detailed analysis and historical data of each API we track, even down to the individual performance from each public cloud data center.